		FeedBack \| Contact Us \| Links \| Site Map

Welcome to Journal of Theoretical and Applied Information Technology

Submit Paper / Call for Papers

Journal receives papers in continuous flow and we will consider articles from a wide range of Information Technology disciplines encompassing the most basic research to the most innovative technologies. Please submit your papers electronically to our submission system at http://jatit.org/submit_paper.php in an MSWord, Pdf or compatible format so that they may be evaluated for publication in the upcoming issue. This journal uses a blinded review process; please remember to include all your personal identifiable information in the manuscript before submitting it for review, we will edit the necessary information at our side. Submissions to JATIT should be full research / review papers (properly indicated below main title).

Journal of Theoretical and Applied Information Technology
October 2024 | Vol. 102 No.19

Title:	CYBERSPACE FOR DETECTING ATTACKS IN AUTONOMOUS VEHICLES BASED APPROACHES
Author:	MOHAMMED Y. ALZAHRANI
Abstract:	With the advancement of technology, cities have progressively grown more intelligent. Smart mobility is a vital component of smart cities, and autonomous vehicles play a fundamental role in enabling smart mobility. Nevertheless, the presence of vulnerabilities in autonomous cars might have a detrimental impact on both the overall quality of life and the safety of individuals. Consequently, several security researchers have examined both offensive and defensive strategies against autonomous cars. Machine learning (ML) and deep leaning (DL) is used in these mobile robots to automate repetitive driving chores and make judgments based on their understanding of the scenario. This research presents showcases the utilization of adversarial instances in connected autonomous vehicles (CAVs) to illustrate how adversarial ML and DL techniques are extremely to detect CANs attacks. The CAVs security system was developed using a dataset acquired from standard research. The dataset includes five types of attacks along with normal packets. The decision tree (DT), extra tree, and Gated Recurrent Units (GRUs) were utilized to identify cyber CAVs threats. The empirical data indicate that the DT technique produced a 99% accuracy rate, while the extra tree and GRU achieved 98% and 96% respectively. Technology demonstrates potential in safeguarding vital infrastructure through the analysis of adversary methods. With near-perfect precision, the performance of all the models constructed in this manner outshone that of previously used models. When it comes to in-vehicle networks (IVN) security, the created system is up to the task.
Keywords:	Cybersecurity, Decision Tree, Gated Recurrent Units, Autonomous Vehicle
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	ENHANCING BANKING SERVICES THROUGH SMART DATA ANALYTICS FRAMEWORK
Author:	YASMINE E. EL-GEMEIE, MOHAMED ABDELSALAM, IBRAHIM F. MOAWAD
Abstract:	Banking targeted marketing strategies have undergone evolution with the integration of predictive analytics and machine learning techniques which play a pivotal role in engaging customers and enticing them to subscribe to various packages and fixed-term deposit offers. The core problem identified was imprecise customer segmentation, resulting in less accurate predictions. The present study focuses on increasing banking sales by predicting customer reactions accurately, contributing to personalized interactions, nurturing customer relationships and proposes the development of a prediction model using machine learning algorithms offering predictive capabilities for sales, customer preferences, new client identification, and efficiency gains. The methodology encompasses data exploration, visualization, preprocessing techniques are applied and implementation of various machine learning models, including XGBoost, Random Forest, Decision Tree, KNN, Logistic Regression, and Naive Bayes. Using a large dataset from a Portuguese bank from Kaggle are employed which used for detection of customer reactions to fixed-term deposit subscriptions. The present results demonstrate high accuracy rate of 93.48% using Random Forest and 92.06% using XGBoost compared to other studies. The consistency observed in cross-validation suggests the models' robustness, emphasizing their suitability for real-world banking campaigns, enhance customer segmentation, optimize targeting, recommend suitable products and improve overall efficiency. While the results are promising, future work should focus on hyperparameter optimization and further refinement of ensemble techniques to boost predictive accuracy.
Keywords:	Artificial intelligence, Machine learning, Banking services, Bank marketing, Decision tree, K-nearest neighbors’ algorithm, Naive Bayes, Support vector machines.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	INTELLIGENT BROKER FOR ERP CLOUD SYSTEM USING MACHINE LEARNING
Author:	AHMED YOUSEF, ESRAA ELHARIRI, MOHAMED H. IBRAHIM
Abstract:	As any business grows, the workload of its administrators also grows exponentially and the need to monitor all parts of the business to ensure efficiency increases. Initially, companies will work with spreadsheets and email, and eventually the boredom of manual logistics will affect their losses. That creates the need for a new smart alternative to help administrators. And from here the light was shed on the enterprise resource planning as an ideal solution. Enterprise resource planning led to the ability of administrators to monitor growth and facilitate work for employees, thus obtaining more clients and more growth. Once that need has been raised, many software companies began to make many programs to fill this gap, and this resulted in giant programs such as (Oracle, SAP, Microsoft, etc.). This need created an urgent necessity, namely, how to choose the appropriate software capable of providing the software to the client in the required manner. The proposed approach explored the application of a multicriteria decision making method (MCDM) that is best worst method (BWM) with k-nearest neighbors (K-NN) algorithm which is one of machine learning (ML) technique used for the evaluation of various enterprise resource planning (ERP(software. First, the BWM model is applied to calculate the weights of criteria, and then, the obtained weights are used in the K-NN method for getting the best alternative. This study will help decision-makers select the best ERP software for all various industries.
Keywords:	MCDM, ML, BWM, K-NN, ERP
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	COMPARATIVE ANALYSIS AND HOW EFFICIENT DEEP LEARNING METHODS OF MALWARE DETECTION
Author:	FIRAS SHIHAB AHMED, NORWATI MUSTAPHA, NOR FAZLIDA MOHD SANI HEAD, RAIHANI MOHAMED
Abstract:	Due to the massive interconnectivity among Internet devices in the Internet of Things (IoT), this led to security challenges in confronting attacks by malware. Detecting malware attacks in the IoT environment is considered a crucial matter that constitutes a challenge for researchers to contribute an accurate method to build a protection system capable of providing security for existing applications in the IoT environment. Today, most of the current research explores deep-learning methods for malware detection. This paper presents an approach that includes analysis to compare the performance of deep learning methods based on opcode in detecting malware in IoT. Four deep learning methods which include Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Gated Recurrent Unit (GRU) are evaluated and compared for accuracy, precision, recall, and F-measure. The idea of this study is based on pre-processing and feature selection by identifying outlier values inside opcodes using the Interquartile range (IQR) technique. Then, the Recursive Feature Elimination (RFE) method has been applied to determine the important features and the suitable hyperparameters to reduce memory space. There are two data sets used in this study to evaluate the performance of the deep learning methods. The first dataset is generated by an IoT-based application with two classes which is considered smaller size than the second dataset which comprises nine different classes. The experimental results showed that the performance of the LSTM method outperformed compared to the other methods which were based on methods for measuring performance and reliability such as accuracy, precision, recall, and F-measure for both data sets. Moreover, used result of receiver operating characteristic (ROC) curves and precision-recall (PR) curves confirm that LSTM is the best method to detect malware. These results will be used as reference results to address the weaknesses of each deep learning method.
Keywords:	Malware Detection, Deep Learning, , AI Methods, Efficiency
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	MODELING OPTIMAL MLP NEURAL NETWORKS WITH DATA MINING FEATURE SELECTION IN CLASSIFYING LUNG CANCER PATIENTS
Author:	AGUS WAHYU WIDODO, TITIS HANDAYANI, FATMA AGUS SETYANINGSIH, IMAM NURHADI PURWANTO, RATNO BAGUS EDY WIBOWO, SAMINGUN HANDOYO
Abstract:	In building a machine learning model, the quality of the model input has a significant role in producing the model with satisfactory performance. This study deploys the data mining feature selection to acquire the independent predictor as the input of multilayer perceptron neural networks (MLP NN) with tuning hyperparameters: node number in the hidden layer and the L2 penalty regularization. The complete predictor dataset is used to build the optimal MLP NN benchmark. Both MLP NN models have the same L2 penalty regularization of 0.05, whereas the node number in the hidden layer of 12 and 7 respectively for the dataset with the complete predictor and independent predictor. The performance evaluation of both MLP NN models in the testing data employing three metrics: accuracy, Mathew's Correlation Coefficient (MCC), and Area under curve (AUC) shows that the optimal MLP NN with independent predictors is not only producing the simpler model but also performing a slightly better than the optimal MLP NN with complete predictors.
Keywords:	Chi-square test, Cross-entropy loss, Data mining, Neural Networks, Regularization method
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	LEVERAGING DEEP LEARNING MODEL WITH OPTIMIZATION ALGORITHM FOR HIGH UTILITY ITEMSET MINING IN TRANSACTIONAL DATA
Author:	G.N. SOWJANYA, M. Babu Reddy
Abstract:	Mining patterns with higher utilization (or higher-utility itemset mining, HUIM) are believed the main problem in the past few decades particularly in the market (For example, supermarkets) engineering later exposes useful products or information for decision-making. Most of the current works concentrated on mining higher-utility itemsets from datasets that showed a very large number of patterns. These processes can’t make accurate decisions in a limited time, for example, real and online decision-making methods, since it is not an unimportant task to discover useful and appropriate information from an enormous quantity of the revealed information in a short time. Mining closed patterns with higher utilization (or termed closed higher-utility pattern mining) is another method to expose concise and smaller patterns with higher utilization in market engineering. Nevertheless, various previous works considered the comprehensive mining growth of all HUIs and they do not reflect the connection between transactions hence once the transactions are not very relevant, the trained method cannot be fully employed for the prediction, which indicates unsuitable results in machine learning (ML) tasks. This manuscript develops a new Deep Learning Based Mining High Utility Itemsets Using the Kepler Optimization Algorithm (DLMHUI-KOA) method. The major objective of the DLMHUI-KOA technique lies in the DL model and optimization parameter using mining high utility itemsets. Initially, the DLMHUI-KOA approach takes place when the design of the recurrent neural network (RNN) model is exploited. Next, the Gated Recurrent Unit (GRU) technique has been applied for predicting the MHUI. At last, the Kepler optimization algorithm (KOA) can be employed for the parameter tuning of the DL model. To determine the higher performance of the DLMHUI-KOA approach, a broad variety of experimentations occurs and the outcomes are examined under several measures such as number of candidates and high utility itemsets, memory usage, and runtime. The comparative analysis reported the improvement of the DLMHUI-KOA approach with recent methods.
Keywords:	High Utility Itemsets; Recurrent Neural Network; Kepler Optimization Algorithm; Gated Recurrent Unit; Memory Usage
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	EMUBOOST OPTIMIZED ENSEMBLE MODEL FOR ROBUST UNDERWATER IMAGE CLASSIFICATION
Author:	SARAVANAN P, VADIVAZHAGAN K
Abstract:	Underwater image classification faces substantial challenges due to low visibility, varying light conditions, and high noise levels. The complexity of underwater environments, characterized by poor contrast and significant distortion, complicates accurate object identification and classification. This research introduces EMUBOOST, an optimized ensemble model that integrates EMU Bird Optimization (EO) with AdamBoost to address these challenges. EMUBOOST leverages the adaptive capabilities of EO to enhance the robustness and efficiency of ensemble learning, improving classification accuracy in underwater environments. The study comprehensively evaluates EMUBOOST's performance in handling the unique difficulties underwater imagery presents. By demonstrating the superiority of this approach in terms of adaptability and accuracy, this research contributes to advancing underwater image analysis with applications in marine biology, underwater surveillance, and environmental monitoring. The findings suggest significant potential for EMUBOOST to improve the reliability and effectiveness of underwater image classification tasks.
Keywords:	Underwater Image Classification - Ensemble Methods - EMUBOOST - AdamBoost - EMU Bird Optimization - Model Optimization - Image Analysis - Classification Accuracy
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	THE ROLE OF ARTIFICIAL INTELLIGENCE IN THE CYBERSECURITY SYSTEM OF BANKING INSTITUTIONS IN THE CONDITIONS OF INSTABILITY
Author:	MAKSYM DUBYNA, ROMAN SHCHUR, OLENA SHYSHKINA, IRYNA SADCHYKOVA, OLENA PANCHENKO, OLENA BAZILINSKA
Abstract:	The development of banking institutions in modern conditions is an important condition for the national economy development. The purpose of the article is to study the possibilities of using artificial intelligence technology to increase the efficiency of the functioning of banking institutions in conditions of instability of the external environment. The article substantiates that stability and favorable conditions for conducting business in the field of banking services are necessary to ensure stable economic development of the country. At the same time, effective functioning of the cybersecurity system of commercial banks plays an important role. The essence of such a system and the peculiarities of its formation and functioning are considered within the scope of the article. It was also established that digitalization of the financial services sector is an objective trend in the development of financial institutions. Accordingly, the use of digital technologies for banks today is extremely important to ensure their competitiveness. Artificial intelligence is one of the digital technologies that opens up new opportunities for banking institutions. At the same time, this technology can both contribute to and create risks for the development of commercial banks. Accordingly, such advantages and disadvantages of its use are considered in the article. Considerable attention is paid to the study of the role of artificial intelligence in ensuring stable functioning of the cybersecurity system of commercial banks. The article also substantiates the lack of a comprehensive understanding of the potential of the artificial intelligence technology for the development of banking among scientists and practitioners, and therefore the application of this technology will only expand to various areas of the operation of banking institutions, including the creation of new algorithms and information innovations to increase stability of the cybersecurity system of these institutions.
Keywords:	Banking Institution, Artificial Intelligence, Cybersecurity, Cyber Risks, Macroeconomic Instability, Information System, Analytical Information.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	ENHANCING NAVIGATION FOR THE VISUALLY IMPAIRED THROUGH OBJECT DETECTION AND 3D AUDIO FEEDBACK
Author:	MONCEF AHARCHI, M HAMED AIT KBIR
Abstract:	Assisting individuals with visual impairments in navigating their surroundings using technological equipment remains a challenging task due to challenges regarding movement, item and person identification, and engagement with the environment. Typically, these devices integrate sensors, visual mechanisms, and tactile or auditory feedback. This article proposes a vision system integrated with 3D audio feedback to improve navigation for the visually impaired people by providing a more intuitive knowledge of object placements along a path by modifying headphone sound level. This system consists of three primary components: firstly, depth calculation utilizing stereoscopic vision to generate a depth map; secondly, object recognition employing a YOLO neural network (CNN) for identifying common objects and Aruco tags for less common ones; and finally, the production of 3D audio based on the depth map and object locations. Subsequently, the user utilizes this spatial audio signal to navigate effectively. When an object is selected using voice commands, the system spell the detected objects names to provide users with direction and distance guidance. During real-world testing, this system has proven to be very helpful and precise in assisting visually impaired people with their navigation.
Keywords:	Visually Impaired, Navigation, Object Detection, 3D Sound, Stereoscopic Vision, Computer Vision, Neural Networks
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	FROM CONTEXT-INDEPENDENT EMBEDDING TO TRANSFORMER: EXPLORING SENTIMENT CLASSIFICATION IN ONLINE REVIEWS WITH DEEP LEARNING APPROACHES
Author:	KOMANG WAHYU TRISNA , JINJIE HUANG , HESHAN LEI , EDDY MUNTINA DHARMA
Abstract:	The exponential progress in technology and the internet has resulted in an unparalleled surge in online engagement, where individuals openly express their viewpoints. Users provide a variety of opinions on politics, events, and product evaluations. User views wield substantial influence on decisions made by both companies and individuals. Manual procedures for identification become impracticable due to the large number of user opinions. Sentiment analysis techniques are employed as a resolution. Deep learning methods have demonstrated potential in accurately predicting polarity from internet reviews, outperforming standard models. Utilizing word embedding techniques in conjunction with deep learning models is crucial for attaining superior results in sentiment classification within the realm of natural language processing (NLP). Furthermore, word embedding approaches like Word2Vec and FastText are thoroughly analyzed for the purpose of mapping text to vectors composed of real numbers. In this study, every assessed deep learning model is combined with both context-independent word embedding and transformer-based embedding. The evaluation of the five model, each utilizing one of the five feature extraction approaches, is conducted using three datasets from distinct domains: IMDB, Amazon, and Yelp. The evaluation is based on multiple metrics, including accuracy, recall, precision, F1-score, and MCC.
Keywords:	Sentiment Analysis, Online Review, Deep Learning, Word Embedding, Transformers
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	ENHANCING ACADEMIC PERFORMANCE PREDICTION FOR AT-RISK STUDENTS: COMPARATIVE ANALYSIS OF MACHINE LEARNING ALGORITHMS IN EARLY WARNING SYSTEMS
Author:	ZAINAB MAHMOUD , ABDELRAZEK WAHBA SAYED
Abstract:	Academic institutions increasingly leverage technology to enhance student performance, particularly through early warning systems that identify at-risk students. These systems utilize various academic and non-academic factors, including grades and attendance, to forecast performance. This study employs a core dataset from Jeddah International College, consisting of 224 instances and 19 attributes, to evaluate the predictive power of several machine learning algorithms. We conduct a comparative analysis of Gaussian Process, Decision Trees, Linear Regression, Ridge Regression, Gradient Boosting, Random Forest, Support Vector Regression, AdaBoost, and LASSO Regression to rank their performance in identifying at-risk students. Our findings reveal that the Gaussian Process and Decision Trees demonstrate the highest predictive capabilities, achieving the highest R² value (0.9657) and the lowest error metrics (RMSE: 0.0424, MSE: 0.0018, MAE: 0.0149). This research outlines the criteria for selecting the most effective models to support academically struggling students.
Keywords:	Machine Learning Algorithms; Academic Metrics; Early Warning System; Students' Performance Prediction.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	A NOVEL META-MODEL BASED ENHANCED LEARNER FOR PREDICTION OF BEHAVIOR TRAITS OF INDIVIDUALS
Author:	CHRISTY JACQUELINE, Dr. K RANJITH SINGH
Abstract:	A person's personality can provide information about their behavior, mental health, emotions, choices in life, interpersonal nature, and ways of thinking. Personality traits are pivotal in identifying behavioral disorders. This paper aims to develop a novel meta-model-based classification algorithm to predict the behavioral characteristics of individuals based on the five-factor model. This proposed work is comprised of two stages, initially, the dataset is trained by the candidate learner, Support vector machine with ensemble learning is used as the base classifier. With the acquired knowledge of the base learner, in the second stage, the super learner known as the meta-model is used for understanding the behavioral disorder of an individual by applying an out-of-fold prediction strategy. The performance analysis of the proposed algorithm-enhanced learner is compared with various algorithms including conventional classifiers, boosting models, bagging models, and ensemble learners. The simulation results prove that the proposed enhanced learner achieves the highest accuracy rate of 0.97%, and precision and recall rates are 0.96% and 0.97% respectively.
Keywords:	Behavioral traits, Base Learner, Candidate Learner, Metamodel, Bagging, Boosting, Out-of-fold
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	HYBRID AUTHENTICATION: CNN-BILSTM KEYSTROKE DYNAMICS AND HOUGH-BASED FINGERPRINT VERIFICATION
Author:	S. RENUKA, N. SURESH KUMAR
Abstract:	This study introduces a hybrid authentication approach integrating Convolutional Neural Networks (CNN) with Bidirectional Long Short-Term Memory (BiLSTM) networks for keystroke dynamics and Hough-based fingerprint verification. The CNN-BiLSTM model leverages deep learning to analyze typing patterns, capturing temporal dependencies and variations in keystroke dynamics. This model is designed to accurately identify users based on their unique typing rhythms, including typing speed and pressure variations. Complementing this, the Hough-based fingerprint verification employs advanced image processing techniques to analyze fingerprint ridge patterns and minutiae with high precision. Ridge lines can be identified and improved with the use of the Hough Transform, and the precision of alignment can be improved with the contribution of subpixel motion estimation. The hybrid approach combines the strengths of behavioral analysis and biometric verification, aiming to provide a robust and reliable authentication mechanism. The integration of CNN-BiLSTM with Hough-based fingerprint verification offers improved security through multi-layered authentication, addressing potential vulnerabilities and enhancing user identity validation in various applications. The implementation of this strategy to provide a solution that is both more secure and accurate for critical systems and applications.
Keywords:	Multi Factor Authentication, Hough Transform, Bidirectional Long Short -Term memory (Bi-LSTM), Convolutional Neural Networks(CNN).
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	A HYBRID CLASSIFIER MODEL - DR-XA FOR DEFECT PRIORITIZATION
Author:	R.ADLINE FREEDA, DR.P.SELVI RAJENDRAN
Abstract:	Fault prioritization in software testing involves determining the sequence in which identified faults should be addressed. Effective fault prioritization is crucial in software development and testing as it helps allocate resources efficiently and ensures that the most critical issues are resolved first. The criteria for prioritization may vary so by addressing the most serious flaws promptly and allocating resources effectively, software quality can be significantly enhanced. Machine learning algorithms offer powerful tools for fault prioritization by leveraging the complexity of the problem and the available data. Common machine-learning approaches for prioritization include classifier models such as Decision Trees, Random Forests, and XGBoost. This research compares the performance of these different classifier models with the proposed DR-XA hybrid prediction model. The DR-XA model, which incorporates advanced techniques for handling unbalanced data and improving prediction accuracy, has been evaluated in the context of fault prioritization. The experimental analysis demonstrates that the DR-XA hybrid model surpasses existing classifier models in prioritization accuracy, achieving superior results compared to current prioritization techniques.
Keywords:	Defect Prioritization, Software Testing, Hybrid classifier model, Prediction, Machine learning
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	ENHANCING NODE SENSING AND AGGREGATION EFFICIENCY IN WIRELESS SENSOR NETWORKS (ENSA)
Author:	JEYA RANI D, NAGARAJAN MUNUSAMY, EZHILARASI M
Abstract:	Wireless Sensor Networks (WSNs) are crucial for various applications, relying on efficient mechanisms for node sensing and data aggregation to optimize energy consumption and prolong network lifetime. Node sensing involves collecting data from sensors within each node, covering diverse environmental parameters like node location, behavior, and history. In this research we proposed ENSA algorithm for node sensing and aggregation. Working at many network levels to reduce data traffic, save energy, and allow in-network processing, data aggregation compiles and summaries data from several nodes before transmission. This work investigates the Hybrid Energy-Efficient Distributed (HEED) clustering and Compressed Aggregation with Correlation (CACC) two main approaches into a new method called Energy-Efficient Node Sensing and Aggregation (ENSA). Aiming for balanced energy usage and long-spanning networks, HEED uses a clustering-based strategy using residual energy and node proximity to elect cluster heads. CACC compresses aggregated data via data correlation between adjacent nodes, therefore lowering transmission overhead and guaranteeing data integrity. This work shows the efficiency and advantages of ENSA, generated from HEED with CACC, using simulations and analysis. The combined approach significantly enhances WSN performance in various application scenarios. It optimizes energy consumption, bandwidth utilization, and data integrity, addressing critical challenges in WSNs. The investigation of node sensing and aggregation techniques such as HEED with CACC underscores their pivotal role in WSNs.
Keywords:	Data Aggregation, Energy Consumption, Network Lifetime, Node Sensing, Wireless Sensor Networks
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	MACHINE LEARNING-BASED CLASSIFICATION AND PREDICTION OF STUDENT STRESS LEVELS: A COMPARATIVE STUDY OF ALGORITHMS
Author:	G. DEENA, A. SANDHYA, K. RAJA
Abstract:	Stress is an increasing issue because of its adverse impacts on students' academic achievement, psychological health, and general well-being. The current research examines the classification and prediction of student stress levels with machine learning algorithms. The objective is to create models that can precisely classify and predict stress levels using a well selected dataset intended to capture elements contributing to student stress. Given the global increase in student stress levels, it is imperative that we deal with this issue to avert serious consequences. This work proposes a machine learning algorithm that measures students' stress levels in classroom environments by leveraging recent advances in computer science, especially in the field of healthcare. The examination examines personal attributes and significant stressors, encompassing academic demands, psychological states, and social engagements. Several machine learning algorithms—Support Vector Machines, Logistic Regression, Naive Bayes, Decision Trees, and Random Forest—are examined, with Naive Bayes identified as the most effective model. It attains a prediction accuracy of 90% and an F1 score of 90%, excelling other classifiers. The model accurately predicts stress levels, enabling early intervention in educational settings, demonstrating the potential of machine learning for monitoring mental health in students.
Keywords:	Stress Level, Machine Learning, Mental Health, Classifiers, Naïve bayes, Logistic Regression
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	DOES E-TRUST AND E-SATISFACTION HAVE AN INFLUENCE IN BUILDING E-LOYALTY TOWARD MOBILE FOOD DELIVERY SERVICES: BASED ON STIMULUS-ORGANISM-RESPONSE THEORY
Author:	WANDA WANDOKO, ASEP NUHDI, IGNATIUS ENDA PANGGATI
Abstract:	The objective of the research is to examine the impact of e-trust and e-satisfaction on the loyalty of mobile food delivery services or MFDS users. The methodology in the study follows quantitative method, with MFDS users in Indonesia as the population. This study collects 651 responses through online surveys which conducted from April 2022 until August 2022. The collected data were analysed with SEMPLS using Smart PLS. The result revealed that direct significant relationship between e-trust and e-satisfaction with e-loyalty. The result also showed that information quality, visual design and navigational design have significant impact on both e-trust and e-satisfaction, while both e-trust and e-satisfaction have influence on building loyalty. This study is one of the few papers to investigate the influence of both e-trust and e-satisfaction on customer e-loyalty toward mobile food delivery services based on SOR theory. This research has several theoretical and managerial implications that can be useful for practitioners such as marketers and MFDS brand owners.
Keywords:	E-Trust, E-Loyalty, E-Satisfaction, Visual Design, Navigational Design, Information Quality.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	TRANSFORMER-BASED MODEL WITH CNN AND CAPSNETS TO IMPROVE MALAY HATE SPEECH DETECTION IN TWEETS
Author:	NUR UMAIRA ABD RAHIM, NORWATI MUSTAPHA
Abstract:	With the rise of social media, the spread of hate speech poses a significant threat to online harmony, especially within the Malay-speaking community. Existing research mainly focuses on high-resource languages like English, leaving a gap in effective HSD for low-resource languages like Malay. Even with a study done in previous research on Malay HSD, there is some room for improvement, and the lack of diverse datasets may significantly affect the system’s overall performance and generalization. Thus, this study proposes a model that uses a transformer-based model named RoBERTa integrated with CNNs and Capsule Networks. RoBERTa is very effective in handling contextual information in bidirectional ways. Experimental results demonstrate that the proposed models, which are RoBERTa, outperform other models in a new dataset in terms of F1-score and accuracy, which are 84.54% and 84.45%, respectively and also outperform the existing dataset, which is 77.67% and 77.45%, respectively. By offering an extensive architecture, this research not only advances the technological area but also tackles social problems by enabling safer online environments for Malay speaker’s communities. Additionally, this research contributes a valuable new Malay Hate Speech dataset, enriching resources for low-resource languages. The results underscore the importance of dataset diversity and advanced NLP techniques in generalizing well across different datasets, making this model practical for real-world applications. Furthermore, this study highlights the global potential of these techniques for improving HSD in other low-resource languages.
Keywords:	Hate Speech Detection, Transformer, Natural Language Processing, XLNet, BERT, RoBERTa
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	MULTICLASS MEMBRANE GASH UNCOVERING AND TAXONOMY USING AMALGAM FEATURES SELECTION BASED ON DEEP CNN
Author:	P. VENKATA RAMA RAJU, MRS. J. HYMAVATHI, ANE ASHOK BABU, SAJJA RADHARANI, YALANATI AYYAPPA, S SINDHURA
Abstract:	One of the primary steps toward clinical therapy may be doing an appropriate sickness analysis. In summary, the area of dermatology is arguably one of the most unpredictable and demanding. Dermatologists always need more patients in order to get the correct conclusion because, all things considered, skin injuries are a severe condition that can affect individuals. For instance, it is essential to have astute frameworks for analysing skin malignant growth early on and, more specifically, to identify and prioritize skin injuries. Subtypes of skin sores that are generally referred to as multiclass skin injuries include Basal Cell Carcinoma (BCC), Melanocytic Nevus (NV), Melanoma (MEL), Actinic Keratosis (AK), Harmless Keratosis Injury (BKL), Squamous Cell Carcinoma (SCC), Dermatofibroma (DF), and Vascular Sore (VASC). The numerous skin injuries and their high likenesses make the multi-class groups still a difficult task. To physically distinguish various skin lesions from dermoscopy photographs, a significant amount of investment and expenditure is required. Therefore, it is crucial to develop computerized diagnostics techniques that can more accurately classify various types of skin lesions. Subsequently this review presents Multiclass skin injury recognition and order using mixture highlight determination in light of Profound Convolutional Brain Organization (DCNN). The presentation of the design is evaluated based on its awareness, accuracy, and explicitness.
Keywords:	Skin lesion, Hybrid structures selection, DCNN,, Dermatology
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	ARTIFICIAL INTELLIGENCE IN THE MECHANISMS OF STATE MANAGEMENT OF THE DEVELOPMENT OF THE REGIONAL AGRARIAN SECTOR IN THE CONTEXT OF THE NATIONAL SECURITY
Author:	IRYNA KOSACH, VIKTORIIA MARHASOVA, VOLODYMYR LAGODIIENKO, OLEKSANDR HOLUBIEV, MAXIM PILIPENKO, ARTEM DEGTYAREV
Abstract:	The purpose of the article is to develop conceptual provisions for improving the mechanisms of state management of the development of the regional agrarian sector based on the artificial intelligence technologies. Artificial intelligence is changing the paradigm of managing national, in particular economic and food security, as it can analyze large sets of data, learn from them and make decisions based on complex patterns. The use of the artificial intelligence technologies can overcome the limitations of the traditional systems of security management, including at the regional level, and improve existing mechanisms of state management. In its expanded form, artificial intelligence in the public administration mechanisms is data analytics, controlled machine learning, and cloud technologies. It is important to coordinate the timing of the use of artificial intelligence in public administration and enterprise management, since digital transformation in public administration is often slower than at the micro level improvement of the existing mechanisms of state management of the development of the agrarian sphere based on artificial intelligence has a two-vector implementation of priority directions. First, the goal of improving the mechanisms of state management is the integration of artificial intelligence into the security management system of enterprises. This concerns, first of all, the provision of cyber security and environmental security to improve threat detection and risk prevention. Based on the use of artificial intelligence, advanced analysis capabilities can be used, ensuring resilience of the public administration system to cyber threats. Specific intelligence technologies should also be selected and configured to eliminate cyber threats to regulate the development of the agrarian sphere, integration with existing systems into the regional system of the agricultural production development.
Keywords:	Artificial Intelligence, Agricultural Sphere, Agro-Industrial Production, State Policy, Mechanisms Of State Management, Regional Policy, Digitalization, Cyber Security, National Security.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	TRAFFIC STATE PREDICTION IN PARIS: LEVERAGING MACHINE LEARNING FOR EFFICIENT URBAN MOBILITY
Author:	ISMAIL ZRIGUI, SAMIRA KHOULJI, MOHAMED LARBI KERKEB, ZINEB REMCH, SALMANE BOUREKKADI
Abstract:	This paper examines the feasibility of applying machine learning models to predict traffic states in Paris, drawing up on real-world data from permanent sensors provided by the City of Paris' Roads and Transport Department. Four popular machine learning models—logistic regression, decision tree, random forest, and k-nearest neighbors—were investigated and evaluated without hyper parameter tuning, revealing insightful performance trends. The analysis delves into the impact of traffic features and missing data, illuminating model strengths and limitations in a practical setting. The paper further explores the potential of innovative approaches involving temporal feature extraction, the use of deep learning models (MLP), and hybrid model combinations with traditional macroscopic traffic models, outlining opportunities for enhancing predictive accuracy.
Keywords:	Imputation, K-nearest neighbors, Logistic regression, Machine learning, Performance evaluation, Random Forest, Traffic prediction, Urban mobility
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	DYNAMICS OF PUBLIC OPINION CHANGE UNDER THE INFLUENCE OF INFORMATION CAMPAIGNS IN ONLINE MEDIA UNDER MARTIAL LAW
Author:	LEONID NOVOKHATKO, TAISIIA BIELOFASTOVA, LIUDMYLA KONONENKO, KATERYNA BARANOVA, ANDRII SINKO
Abstract:	The development of an information society contributes to the formation of online media, which affects public opinion. The article aims to study the specifics of public opinion change under the influence of information campaigns in online media under martial law. For this, methods of comparative analysis, calculations of the weighting factor, the advance coefficient, and the Mann-Whitney coefficient were used. The study found that 'HROMADSKE', 'Suspilne Movlennia', and 'ZN.UA' are the most popular online media in Ukraine, which contribute to the presentation of trustworthy and relevant information. The most spread world online media are 'Associated Press', and 'The New York Times', which reflect various themes, and promote the engagement of a larger number of readers. The most influential parameters affecting public opinion were found to be the filling with the content according to referring to a certain source. They also include information substantiation with relevant facts, daily information coverage, information analysis by competent individuals, information sharing by different sources, and forms of presenting information. Analysis of the influence of information campaigns showed that they contributed to the change of public opinion concerning the government (31%), and attitude to the government of other countries (25%). The practical value of the work lies in studying the factors positively affecting public opinion change, which may be considered in journalists training. Study perspectives may be related to the determination of the effect of informational opinion on the influence of citizens of different age groups.
Keywords:	Information Source, Political Situation, Information Trustfulness, Forms of Presenting Information, Issue Publicity
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	SMARTCHAIN: ENHANCING IOT SECURITY AND TRUST MANAGEMENT USING BLOCKCHAIN FOR REAL-TIME, DECENTRALIZED APPLICATIONS
Author:	HARI PRASAD CHANDIKA, Dr KONTHAM RAJA KUMAR
Abstract:	As more IoT devices are being developed, the demand for safe and reliable applications also increases. However, trust and security between IoT ecosystems is difficult because their nature of distributed system and insecure property in the world. This paper suggests SMARTCHAIN as a new blockchain-oriented trust management framework to solve major concerns in IoXs. SMARTCHAIN has incorporated different parameters (system usage, number of transaction requests, the number of nodes data rate received or transferred by a server from IoT Device sensory Data node and computational time in which given server takes to initiate verifications) unlike previous works into trust assessment mechanism for obtaining specific level assessment on an individual device connected over network. The blockchain technology is a key that can provide protection of distributed data integrity, immutability and transparency with the combinance of property trust management using decentralized structures effectively lowering computational overheads and response time [18]. Our extensive simulations show that SMARTCHAIN performs better in scale, low-latency and efficiency compared to previous works. Summary This work contributes a novel, scalable and efficient blockchain based solution for trust management in IoT applications, addressing the challenges left uncovered by state-of-the-art approaches.
Keywords:	Blockchain, Data Integrity, Decentralized Trust Management, IoT Applications. Security, Trustworthiness
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	AIRADF: ARTIFICIAL INTELLIGENCE ENABLED CLINICAL DECISION SUPPORT SYSTEM FOR DIAGNOSING RHEUMATOID ARTHRITIS USING X-RAY IMAGES
Author:	MOHD ABDUL RAHMAN, Dr. MANIZA HIJAB, Dr. S. FOUZIA SAYEEDUNNISA, MASARATH SABA, Dr.GOURI R PATIL, Dr.A. C.PRIYA RANJANI
Abstract:	Medical image analysis plays a crucial role in healthcare, particularly in computer vision applications. Artificial Intelligence (AI) has greatly contributed to solving various issues in the healthcare industry, including disease diagnosis and classification. Rheumatoid Arthritis (RA) is an autoimmune disease that causes serious health problems. The current learning-based approaches for RA diagnosis require improvements in pipelining and optimizations. In this paper, we propose a deep learning-based framework called Artificial Intelligence (AI) for RA Diagnosis Framework (AIRADF). This framework includes functionality for preprocessing and training Region of Interests (ROIs) for automatic RA detection and classification. The RA detection process utilizes a deep learning model known as Faster RCNN, while RA classification is carried out by an enhanced UNet model. We introduce an algorithm called Learning Based Rheumatoid Arthritis Detection (LbRAD). Our empirical study using X-ray images demonstrates that the proposed algorithm outperforms many existing deep learning models in RA detection and classification, achieving the highest accuracy of 92.81% and 94.58%, respectively. Additionally, our framework enables multi-class classification beyond RA detection, resulting in a Clinical Decision Support System (CDSS) that can aid healthcare professionals in RA prognosis.
Keywords:	Rheumatoid Arthritis, Deep Learning, Artificial Intelligence, Image Processing, Rheumatoid Arthritis Prognosis
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	OBJECT DETECTION TECHNIQUES FOR STRAWBERRY DISEASE DETECTION : A COMPREHENSIVE REVIEW
Author:	ZINEB REMCH, SAMIRA KHOULJI, MOHAMED LARBI KERKEB
Abstract:	Strawberry, which is a popular fruit, is known for its high content of vitamin C and antioxidants, thus contributing to cardiovascular health and blood sugar control. Faced with the challenges posed by diseases affecting its cultivation, such as anthracnose and powdery mildew, the integration of advanced technologies has become crucial to improve productivity compared to conventional agricultural methods. In recent years, deep learning techniques have been widely used in various fields of computer vision, demonstrating their potential for strawberry disease detection. However, the lack of in-depth discussions on the application of deep learning to this culture highlights the need for a comprehensive review of recent technologies. This article provides a comprehensive review of recent advances in this field and highlights four main models: YOLO, Mask R-CNN, RetinaNet, and SSD, which have been widely used in object detection. It also explores different databases available in the literature, while highlighting the challenges of using them for real-time research.
Keywords:	Strawberry Disease, Object Detection, Deep Learning, Yolo, Mask R-CNN, Retinanet, SSD
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	PIXEL-BASED ALGORITHM IN BRAIN TUMOR CLASSIFICATION MEASURING AREA
Author:	DWI SWASONO RACHMAD, JOHAN HARLAN, MOHAMMAD IQBAL
Abstract:	Magnetic Resonance Imaging (MRI) is a medical computer technology that can be utilized for the diagnosis of brain cancers. MRI is a medical imaging technique that utilizes magnetic fields based on the principles of computed tomography. Nevertheless, there are still deficiencies in the areas of interpretation, temporal analysis, and visualization. The primary objective of this study is to identify the precise features of the Glioma tissue, including its location and size. The study utilized MRI images in bmp format obtained from the Cipto Mangunkusumo Jakarta Central General Hospital. The study focused on developing algorithms using active contour methods, otsu method, and hybrid method. Additionally, the study incorporated object detection ROI combined with region feature methods. The area of the glioma was calculated using Matlab and Python tools throughout the testing phase. This study utilizes imaging data from 13 T1 contrasted axial sequence components. The outcome of the segmentation and extraction method is utilized to compute the glioma's area, which is determined by the pixels that have undergone processing. The hybrid method effectively distinguishes the object under study from other objects. This method achieves an average accuracy rate of 99% when using numerical methods that involve absolute and relative error calculations, as well as extraction segmentation processes with template matching algorithms. The average precision value of +1 indicates that feature extraction with a hybrid approach is dependable for clinical evaluation
Keywords:	Brain Tumor, Hybrid, MRI, Pixel, Template Matching
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	INTEGRATION OF AUTOMATED GRABCUT ALGORITHM WITH DEEPLABV3+ TO ENHANCE IMAGE SEGMENTATION FOR ACCURATE LEAF DISEASE DETECTION AND CLASSIFICATION
Author:	SREYA JOHN, Dr. P. J. ARUL LEENA ROSE
Abstract:	Detection and classification of plant diseases play a significant role in various fields such as plant pathology, agriculture and environmental studies. To produce effective segmentation of leaf images, this study provides a hybrid strategy that blends deep learning approaches with the enhanced GrabCut algorithm. The proposed method automates the original GrabCut algorithm in order to build initial masks, which are then refined using the powerful DeepLabv3+ model. A detailed correlative analysis is also performed to demonstrate that the suggested model with an efficacy of 95.99% outperforms existing deep learning models such as Unet, SegNet, and other commonly used segmentation methods. The results are obtained using evaluation metrics such as pixel accuracy, intersection of union, precision, recall, and F1 score, demonstrating that incorporating deep learning into the GrabCut algorithm significantly enhanced the leaf image segmentation process.
Keywords:	Segmentation, Leaf Disease Detection, Grabcut Algorithm, Deeplabv3+ Network
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	OPTIMIZING ENCRYPTED DATA RETRIEVAL: PARALLEL SEARCH TECHNIQUES WITH HASH TABLE INDEXING
Author:	WAFA ALDABABAT
Abstract:	String matching is one of the fundamental operations in many applications, it is more challenging today to handle string searching while maintaining the security and privacy of data. In this paper a mechanism for searching in encrypted data is presented, this is done by using index-based searching on encrypted data, with using hash tables as index for searching and encrypting both the keys and values stored in the hash table for security. Also, parallel searching of multiple patterns is used in our implementation by using multiple threads is added to enhance efficiency. The results show that the approach is feasible and has a potential application in secure data retrieval systems, because it improves the searching performance while maintaining the security and privacy of data.
Keywords:	String matching, Encryption, Parallel search, Security, Hash tables, Index-Based searching.
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text

Title:	REVOLUTIONIZING E-COMMERCE WITH AI CHATBOTS: ENHANCING CUSTOMER SATISFACTION AND PURCHASE DECISIONS IN ONLINE MARKETPLACE
Author:	TANTY OKTAVIA, CAROLINE WIDIAWATI ARIFIN
Abstract:	This research investigates the impact of chatbot features on consumer satisfaction and purchase decisions within e-commerce platforms. The research identifies key factors contributing to consumer satisfaction, including interactivity, communication manner, responsiveness, and perceived usability. The findings demonstrate that responsiveness significantly influences customer satisfaction, which in turn, strongly affects purchase decisions. Contrary to expectations, perceived usability and interactivity did not show a statistically significant effect on customer satisfaction. These results suggest that while ease of use is important, responsiveness and communication manner may play a more critical role in shaping customer experiences. Effective communication, though not directly linked to higher satisfaction in this study, remains crucial for enhancing the natural and intuitive interaction with chatbots. The study underscores the importance of optimizing chatbot features to create a seamless and enjoyable shopping experience, ultimately boosting consumer satisfaction and encouraging repeat purchases. As the e-commerce sector continues to expand, businesses must focus on enhancing these aspects to remain competitive and meet evolving consumer needs.
Keywords:	E-Commerce, AI Chatbots, Customer Satisfaction, Purchase Decisions, Responsiveness
Source:	Journal of Theoretical and Applied Information Technology 15^th October 2024 -- Vol. 102. No. 19-- 2024
Full Text